Abstract
We adapted a recent computational model of achromatic color induction (Rudd, JoV, 2010; J Electron Imaging, 2017) to model chromatic induction along the L/M color axis. Our equiluminant display contained two 0.7 deg disks, presented 8.5 deg apart and surrounded by annuli of equal but variable width (0.35–4.0 deg). The observer adjusted the L/M cone contrast of the left disk—which was surrounded by a 498.0 L-, 204.0 M-troland annulus—to match the perceived hue of the 448.2 L-, 217.0 M-troland right disk, whose surround annulus varied across trials over the range 443.8–635.3 L and 233.9–140.6 M trolands. Our model assumes that the color of each disk is computed by a weighted spatial summation of logarithmic steps in L/M contrast at the inner and outer annulus edges. Prior to the spatial summation stage of the model, the neural gains applied to edges are adjusted by a contrast gain control acting between edges. The model makes two key predictions: 1) hue matches should vary as a quadratic function of the L/M contrast of the variable annulus when plotted in log-log coordinates; 2) the 1st- and 2nd-order regression coefficients estimated by fitting the quadratic model to matching data from individual observers and annulus sizes will be related to each other by a constant of proportionality that depends only the L/M contrasts of the fixed disk and background (fixed display parameters). A variant of the induction model based on the cone contrast measure logL/log(L+M) obeyed the model predictions to within 2.1% error.